Monitoring salt marshes using high spatial resolution satellite imagery for mapping and change detection: Protocol development for northeast coastal parks

Wang, Q.T.; Campbell, Anthony

In October 2012, Hurricane Sandy imposed huge impacts to the coastal parks in the Northeast United States. Post-Sandy rapid response aerial imageries, collected and published by the Department of Commerce (DOC), National Oceanic and Atmospheric Administration (NOAA), National Ocean Service (NOS) and National Geodetic Survey (NGS), revealed the significant impacts of this superstorm. The images illustrated that the salt marshes in these parks were significantly affected. Fire Island National Seashore (FIIS) alone, for example, was breached at the eastern end and significant overwash across the entire island. The National Park Service (NPS) identified that the loss of salt marsh habitats in coastal parks, in particular for those affected by 2012 Hurricane Sandy, warrants change analysis and development of a cost-effective, high spatial resolution satellite remote sensing-based salt marsh change detection protocol. The protocol developed should be applicable toward a long term salt marsh change analysis and monitoring of the northeast coastal parks managed by the Northeast Coastal and Barrier Network (NCBN) of the NPS Inventory & Monitoring Program. This report details a mapping and change detection protocol using very high spatial resolution satellite remote sensing data. The protocol was developed for the salt marsh mapping and change analysis of Jamaica Bay within the Gateway National Recreation Area (GATE). The protocol was also used for salt marsh mapping at Fire Island National Seashore (FIIS) and Assateague Island National Seashore (ASIS). The report demonstrates a methodology for mapping fine-scale changes in salt marsh extent. Change derived from natural disturbances, restoration, and stressors necessitates continued monitoring to inform management. Remote sensing has seen many advances in the recent years, including the combination of very high resolution (VHR) imagery with Object-based Image Analysis (OBIA), hyperspectral imagery to discriminate individual types of vegetation, LiDAR to improve image classification, and increased prevalence of Synthetic Aperture Radar (SAR) which is particularly suited for wetland mapping (Klemas, 2013). Previous salt marsh mapping of Jamaica Bay in 2003 and 2008 utilized the Quickbird-2 (QB-2) sensor, a VHR commercial satellite, with a traditional pixel-based image classification approach (Wang, Christiano & Traber, 2010; Wang et al., 2007). Image classification methods are constantly expanding and new options have become widespread since the last mapping of Jamaica Bay. This project evaluated QB-2 and Worldview-2 (WV-2) satellite data for mapping and monitoring salt marsh in the Bay. WV-2 has additional spectral and spatial resolution when compared with QB-2 images. In this project, image classifications with pixel-based and OBIA were compared to assess their applicability for salt marsh mapping with VHR satellite imagery. The comparisons justified the shift from pixel-based classification to OBIA due to the finer spatial resolution and availability of ancillary data. Mapping accuracy was assessed with both internal cross-validation and error matrix accuracy assessments detailed in the technical workflow. The accuracy assessment of Jamaica Bay salt marsh mapping was conducted with field knowledge from site visits in 2014 and 2015. Land cover data were collected with a Trimble XH at a horizontal accuracy of < 1m and a GPS enabled digital camera. This knowledge combined with QB-2 and WV-2 satellite and Google Earth imageries were utilized to conduct the accuracy assessments. The assessments were analyzed for overall accuracy, Kappa coefficient, producer’s accuracy and user’s accuracy. Kappa coefficient is a measure of how much a classification improves on a random classification. The 2012 mapping of Jamaica Bay was used to develop the protocol. The highest overall accuracy.......

Monitoring salt marshes using high spatial resolution satellite imagery for mapping and change detection: Protocol development for northeast coastal parks

Series